Dark Object Subtraction Research Articles

A comparison of radiometric normalization methods when filling cloud gaps in Landsat imagery

Mapping persistently cloudy tropical landscapes with optical satellite imagery usually requires assembling the clear imagery from several dates. This study compares methods for normalizing image data when filling cloud gaps in Landsat imagery with imagery from other dates. Over a complex tropical island landscape, namely St. Kitts and Nevis and the island of St. Eustatius, all of the methods tested reduce interdate image differences for ETM+ bands 1-5 and 7, NDVI, and band 4:5 ratio. Regression tree normalization reduces the interdate differences more consistently than linear regression or histogram matching. Normalizing ETM+ images with regression trees can produce more seamless imagery than linear normalization, histogram matching, or image-based atmospheric correction via dark object subtraction. More seamless imagery enhances visual interpretation and helps reveal the distributions of forest formations in these landscapes. Decision tree classification of cloud-filled Landsat imagery can accurately map land cover and detailed forest formations. Decision tree classification accuracy is not highly dependent on the method used to make the cloud-filled imagery, however, at least as long as (i) classification model training data reflect class spectral variability, and (ii) ancillary spatial data that relate to the distributions of classes are used in the classification. Cloud-filled imagery is also known as cloud-cleared imagery.

Radiometric correction of multi-temporal Landsat data for characterization of early successional forest patterns in western Oregon

Detecting and characterizing continuous changes in early forest succession using multi-temporal satellite imagery requires atmospheric correction procedures that are both operationally reliable, and that result in comparable units (e.g., surface reflectance). This paper presents a comparison of five atmospheric correction methods (2 relative, 3 absolute) used to correct a nearly continuous 20-year Landsat TM/ETM+ image data set (19-images) covering western Oregon (path/row 46/29). In theory, full absolute correction of individual images in a time-series should effectively minimize atmospheric effects resulting in a series of images that appears more similar in spectral response than the same set of uncorrected images. Contradicting this theory, evidence is presented that demonstrates how absolute correction methods such as Second Simulation of the Satellite Signal in the Solar Spectrum (6 s), Modified Dense Dark Vegetation (MDDV), and Dark Object Subtraction (DOS) actually make images in a time-series somewhat less spectrally similar to one another. Since the development of meaningful spectral reflectance trajectories is more dependant on consistent measurement of surface reflectance rather than on accurate estimation of true surface reflectance, correction using image pairs is also tested. The relative methods tested are variants of an approach referred to as “absolute-normalization”, which matches images in a time-series to an atmospherically corrected reference image using pseudo-invariant features and reduced major axis (RMA) regression. An advantage of “absolute-normalization” is that all images in the time-series are converted to units of surface reflectance while simultaneously being corrected for atmospheric effects. Of the two relative correction methods used for “absolute-normalization”, the first employed an automated ordination algorithm called multivariate alteration detection (MAD) to statistically locate pseudo-invariant pixels between each subject and reference image, while the second used analyst selected pseudo-invariant features (PIF) common to the entire image set. Overall, relative correction employed in the “absolute-normalization” context produced the most consistent temporal reflectance response, with the automated MAD algorithm performing equally as well as the handpicked PIFs. Although both relative methods performed nearly equally in terms of observed errors, several reasons emerged for preferring the MAD algorithm. The paper concludes by demonstrating how “absolute-normalization” improves (i.e., reduces scatter in) spectral reflectance trajectory models used for characterizing patterns of early forest succession.

Image-based atmospheric correction of QuickBird imagery of Minnesota cropland

High spatial resolution QuickBird satellite data have provided new opportunities for remote sensing applications in agriculture. In this study, image-based algorithms for atmospheric correction were evaluated on QuickBird imagery for retrieving surface reflectance ( ρ λ ) of corn and potato canopies in Minnesota. The algorithms included the dark object subtraction technique (DOS), the cosine approximation model (COST), and the apparent reflectance model (AR). The comparison with ground-based measurements of canopy reflectance during a 3-year field campaign indicated that the AR model generally overestimated ρ λ in the visible bands, but underestimated ρ λ in the near infrared (NIR) band. The DOS–COST model was most effective for the visible bands and produced ρ λ with the root mean square errors (RMSE) of less than 0.01. However, retrieved ρ λ in the NIR band were more than 20% (mean relative difference or MRD) lower than ground measurements and the RMSE was as high as 0.16. The evaluation of the COST model showed that atmospheric transmittance ( T λ θ ) was substantially overestimated on humid days, particularly for the NIR band because of the undercorrection of water vapor absorption. Alternatively, a contour map was developed to interpolate appropriate T λ θ for the NIR band for clear days under average atmospheric aerosol conditions and as a function of precipitable water content and solar zenith angle or satellite view angle. With the interpolated T λ θ , the accuracy of NIR band ρ λ was significantly improved where the RMSE and MRD were 0.06 and 0.03%, respectively, and the overall accuracy of ρ λ was acceptable for agricultural applications.

Retrieval of wheat leaf area index using price approach based on inversion of canopy reflectance model

This paper reports estimation of the Leaf Area Index (LAI) of wheat crop from IRS-LISS-III data using Price (1993) approach. Empirical approach for LAI estimation with different NDVI estimation procedures viz. radiance, apparent reflectance and dark object subtraction (DOS) based atmospheric correction were also evaluated. Validation of LAI retrieval and NDVI normalizations were carried out using field level measurements of crop LAI and spectral property using canopy analyzer and spectro-radiometer, respectively over selected fields in Bhopal District, Madhya Pradesh. It was observed that empirical relations are sensitive to the NDVI estimation approach and DOS method performed better as compared to other two approaches. It was also observed that LAI estimation from Price algorithm is sensitive to the crop attenuation coefficients. Crop specific attenuation coefficients reported in literature for Indian cultivars gave higher accuracy. The root mean square (RMS) error of 0.77 for LAI estimation was achieved using above described approach.

Urban Aerosol Optical Thickness Retrieval from MODIS Data

Aerosol plays an important role in global climate change and remote sense provides a better method to retrieve aerosol optical thickness (AOT) at large scale. Currently dark object subtraction and contrast reduction method are two widely used methods for MODIS data to get AOT. Due to the facts that there are few pixels whose reflectance in urban area are not small enough, the urban aerosols are from multi-sources and their components are difficult to verify, dark object method may lead to big errors if it is used in urban area. Here the different response of MODIS band 7, 3, 1 to aerosol are used together with the ground based measurements, a formula is put forward as τ = 4.4376×ρ_7+50.5579×ρ_3-24.3317×ρ_1-3.5575 to calculate AOT in urban area. The error analysis shows that the error of 81% results is less than 10% and 61% is less than 6%. This formula is derived from the data observed in one day. The more accurate result can be derived from the meteorological data provied by hundreds of weather sations.

Monitoring forest succession with multitemporal landsat images: factors of uncertainty

This study evaluates uncertainty factors in using multitemporal Landsat images for subtle change detection, including atmosphere, topography, phenology, and sun and view angles. The study is based on monitoring forest succession with a set of multiple Landsat Thematic Mapper/Enhanced Thematic Mapper Plus (TM/ETM+) images spanning 15 years over the H.J. Andrews Experimental Forest in the Western Cascades of Oregon. The algorithms for removing atmospheric effects from remotely sensed images evaluated include a new version of the dark object subtraction (DOS3) method, the dense dark vegetation (DDV) method, the path radiance (PARA) approach, and the 6S radiative transfer codes. We found that the DOS3 approach undercorrects the image, and the recently developed DDV and PARA approaches can produce surface reflectance values closely matching those produced by 6S using in situ measurements of atmospheric aerosol optical depth. Atmospheric effects reduce normalized difference vegetation index (NDVI) and greenness, and increase brightness and wetness. Topography modifies brightness and greenness, but has minimal effects on NDVI and wetness, and it interacts with sun angle. Forest stands at late successional stages are more sensitive to topography than younger stands. Though the study areas are covered predominantly by evergreen needleleaf forests, phenological effect is significant. Sun angle effects are confounded with phenology, and reflectance values for stands at different successional stages are related to sun angles nonlinearly. Though Landsat has a small field of view angle, the view angle effects from overlapping Landsat scenes for a mountainous forested landscape may not be ignored when monitoring forest succession with multitemporal images.

Forest mapping with a generalized classifier and Landsat TM data

Monitoring landcover and landcover change at regional and global scales often requires Landsat data to identify and map landscape features and patterns with sufficient detail. Analytical methods based on image-by-image interpretation are too time-consuming and labor-intensive for studies of large areas to be undertaken with any degree of frequency. One potential solution is to develop algorithms or classifiers that can be generalized beyond the arena of the initial training to new images from different spatial, temporal or sensor domains. Building upon earlier success with a generalized classifier to monitor forest change, we now address the question of generalization for classifications of stable landcovers. We evaluate the ability of a supervised neural network, Fuzzy ARTMAP, to identify conifer forest across time and space with Landsat Thematic Mapper (TM) images for a region in northwest Oregon. We also assess the effects of atmospheric corrections on generalized classification accuracies. Using midsummer images atmospherically corrected with a simple dark-object-subtraction (DOS) method, there is no statistically significant loss of accuracy as the classification is extended from the initial training image to other images from the same scene (path and row): temporal generalization is successful. Extending the classifier across space and time to nearby scenes results in a mean decline of 8–13% accuracy depending on the atmospheric correction used. Obvious sources of error, such as seasonality, solar angle variation, and complexity of landcover identification, do not explain the decline in error. Additionally, the patterns in generalization accuracies are complex, and the relationship between pairs of training and testing images is not necessarily reciprocal, i.e., good training data are not necessarily good testing data. Simple DOS atmospheric corrections produce classifications with comparable accuracies as classifications from the more complex radiative transfer corrections. These findings are based on over 200 classifications. A high degree of variability in the classification accuracies underscores the importance of extensive, in-depth analysis of remote sensing techniques and applications, and highlights the potential problem for misleading results based on just a few tests. Generalization is well suited for multitemporal classifications of one Landsat scene. Using simple DOS and midsummer images, generalization offers the opportunity for frequent landcover mapping of a Landsat scene without having to retrain the classifier for each time period of interest. However, at this point, the utility of regional landcover mapping with a generalized classifier remains limited.

A simplified method for retrieval of ground level reflectance of targets from airborne video imagery

This paper describes an efficient method for retrieval of ground reflectance characteristics of targets from calibrated multispectral airborne video data for routine operational airborne missions. The method uses a simplified atmospheric scattering model in combination with a dark-object subtraction procedure to estimate the effect of the atmosphere in the path between the target and the sensor, as well as the adjacent environmental effect, on the radiation signal received by an airborne sensor. The simplicity of the atmospheric scattering model is maintained by the assumption that the air density within the targetsensor path in the lower atmosphere is sufficiently uniform for operations of the Charles Sturt University's (CSU) Multispectral Airborne Video System (MAVS). The MAVS acquires imagery in blue, green, red and near-infrared (NIR) narrow spectral bands. The MAVS is radiometrically calibrated and has a consistent radiometric response in-flight. An important feature of the new method is the coupling of the image based brightness data (DN) of a dark-object and the system radiometric calibration coefficients to determine the path reflectance and the environmental reflectance of the target. The sum of the path reflectance and the environment reflectance is known as haze reflectance. The haze reflectance indicates the amount of atmospheric haze in the airborne imagery. The simplified atmospheric model is then employed to determine the actual ground reflectance of the targets using the haze subtracted apparent (total) reflectance of the target at the altitude of the airborne sensor. The apparent reflectance of the target at the sensor altitude is obtained directly from the image based DN data and the system radiometric calibration coefficients. An interesting aspect of this simplified method is that an estimate of the environmental reflectance can be obtained as a by-product of the atmospheric haze calculation using a dark-object subtraction technique. The retrieved ground reflectance characteristics from calibrated MAVS imagery are now being used routinely for remote quantitative monitoring of agricultural and environmental targets.

Classification and Change Detection Using Landsat TM Data: When and How to Correct Atmospheric Effects?

The electromagnetic radiation (EMR) signals collected by satellites in the solar spectrum are modified by scattering and absorption by gases and aerosols while traveling through the atmosphere from the Earth's surface to the sensor. When and how to correct the atmospheric effects depend on the remote sensing and atmospheric data available, the information desired, and the analytical methods used to extract the information. In many applications involving classification and change detection, atmospheric correction is unnecessary as long as the training data and the data to be classified are in the same relative scale. In other circumstances, corrections are mandatory to put multitemporal data on the same radiometric scale in order to monitor terrestrial surfaces over time. A multitemporal dataset consisting of seven Landsat 5 Thematic Mapper (TM) images from 1988 to 1996 of the Pearl River Delta, Guangdong Province, China was used to compare seven absolute and one relative atmospheric correction algorithms with uncorrected raw data. Based on classification and change detection results, all corrections improved the data analysis. The best overall results are achieved using a new method which adds the effect of Rayleigh scattering to conventional dark object subtraction. Though this method may not lead to accurate surface reflectance, it best minimizes the difference in reflectances within a land cover class through time as measured with the Jeffries–Matusita distance. Contrary to expectations, the more complicated algorithms do not necessarily lead to improved performance of classification and change detection. Simple dark object subtraction, with or without the Rayleigh atmosphere correction, or relative atmospheric correction are recommended for classification and change detection applications.

A simplified method for retrieval of ground level reflectance of targets from airborne video imagery

This paper describes an efficient method for retrieval of ground reflectance characteristics of targets from calibrated multispectral airborne video data for routine operational airborne missions. The method uses a simplified atmospheric scattering model in combination with a dark-object subtraction procedure to estimate the effect of the atmosphere in the path between the target and the sensor, as well as the adjacent environmental effect, on the radiation signal received by an airborne sensor. The simplicity of the atmospheric scattering model is maintained by the assumption that the air density within the targetsensor path in the lower atmosphere is sufficiently uniform for operations of the Charles Sturt University's (CSU) Multispectral Airborne Video System (MAVS). The MAVS acquires imagery in blue, green, red and near-infrared (NIR) narrow spectral bands. The MAVS is radiometrically calibrated and has a consistent radiometric response in-flight. An important feature of the new method is the coupling of the image based brightness data (DN) of a dark-object and the system radiometric calibration coefficients to determine the path reflectance and the environmental reflectance of the target. The sum of the path reflectance and the environment reflectance is known as haze reflectance. The haze reflectance indicates the amount of atmospheric haze in the airborne imagery. The simplified atmospheric model is then employed to determine the actual ground reflectance of the targets using the haze subtracted apparent (total) reflectance of the target at the altitude of the airborne sensor. The apparent reflectance of the target at the sensor altitude is obtained directly from the image based DN data and the system radiometric calibration coefficients. An interesting aspect of this simplified method is that an estimate of the environmental reflectance can be obtained as a by-product of the atmospheric haze calculation using a dark-object subtraction technique. The retrieved ground reflectance characteristics from calibrated MAVS imagery are now being used routinely for remote quantitative monitoring of agricultural and environmental targets.

An atmospheric correction method for the automatic retrieval of surface reflectances from TM images

Abstract Most of the atmospheric correction methods proposed in the literature are not easily applicable in reaJ cases. The most sophisticated models frequently require inputs which are not commonly available, whilst traditional simple dark object subtraction techniques do not generally give real reflectance values. In the present work an atmospheric correction method applicable to Landsat-TM data is described, which requires only inputs that are commonly available and the presence in the imaged scenes of some dark surfaces in TM bands 1 (blue) and 3 (red). The method consists of an inversion algorithm based on a simplified radiative transfer model in which the characteristics of atmospheric aerosols are estimated by the use of the path radiance in two TM bands rather than a priori assumed. On the basis of this information, which is crucial for determining the atmospheric properties, the retrieval of real reflectances from TM images is possible. The method can be applied to all TM scenes in which some dark points can be realistically supposed to be present, which is particularly advantageous in retrospective studies. Several TM scenes taken from different landscapes and in different seasons were corrected using the model. The reflectance values found were tested against ground measurements and compared with data from the literature. The results show a substantial improvement in the accuracy of the reflectance estimates with respect to estimates without atmospheric correction. Given some care in the identification of dark values, the relative error in actual reflectance retrieval is always rather low (≅10–20 per cent); this error can be considered acceptable for most practical applications.

Processing of multitemporal Landsat TM imagery to optimize extraction of forest cover change features

Digital procedures to optimize the information content of multitemporal Landsat TM data sets for forest cover change detection are described. Imagery from three different years (1984, 1986, and 1990) were calibrated to exoatmospheric reflectance to minimize sensor calibration offsets and standardize data acquisition aspects. Geometric rectification was followed by atmospheric normalization and correction routines. The normalization consisted of a statistical regression over time based on spatially well-defined and spectrally stable landscape features spanning the entire reflectance range. Linear correlation coefficients for all bitemporal band pairs ranged from 0.9884 to 0.9998. The correction mechanism used a dark object subtraction technique incorporating published values of water reflectance. The association between digital data and forest cover was maximized and interpretability enhanced by converting band-specific reflectance values into vegetation indexes. Bitemporal vegetation index pairs for each time interval (two, four, and six years) were subjected to two change detection algorithms, standardized differencing and selective principal component analysis. Optimal feature selection was based on statistical divergence measures. Although limited to spectrally-radiometrically defined change classes, results show that the relationship between reflective TM data and forest canopy change is explicit enough to be of operational use in a forest cover change stratification phase prior to a more detailed assessment. >

Chrom: A C program to evaluate the application of the dark object subtraction technique to digital remote-sensing data

This program, written in Turbo-C, version 2.0, permits the user to enter data from several earth-resource satellite systems, determine corresponding radiances (using published calibration data) and plot chromaticity coordinates on the usual chromaticity diagram. If the user enters the digital numbers (DNs) in three spectral bands for a dark object selected by the user, the program will use these values to perform the dark object subtraction technique for adjusting for the effects of atmospheric scattering, and plot chromaticity coordinates for both original and corrected values. The program allows the user to enter values from the screen, or from a file that records several sets of DNs. Calibration data also can be entered by the user, to allow for use with revised calibration data, or with systems other than those discussed here. The program also allows the user to apply a modified dark object subtraction technique as proposed by P. S. Chavez in 1988. The user selects a scattering model, based upon an assessment of prevailing atmospheric conditions, and selection of one of the visible bands to be used as a base for calculation of corrections to be applied to other bands. The chromaticity technique permits the user to examine the spectra of features before and after application of the correction, to assess if the adjustment is satisfactory.

Evaluation of simplified procedures for retrieval of land surface reflectance factors from satellite sensor output

In response to the need for a simple atmospheric correction method and the consequent verification of such a method, an experiment was conducted to acquire a data set suitable for testing atmospheric correction procedures under a variety of atmospheric conditions. Several procedures, including radiative transfer codes (RTCs) with simulated atmospheres, image-based procedures and dark-object subtraction (DOS), were evaluated by comparing surface reflectance factors derived from Landsat Thematic Mapper (TM) digital data with low-altitude, aircraft-based measurements for seven dates over a 1-year period. Acceptable results, approximately ± 0.02 reflectance (1 σ RMS), were achieved based on an RTC with appropriate simulated atmospheres. The DOS technique was the least accurate method and, in fact, produced greater error in estimations of near-IR reflectance than no correction at all. Two hybrid approaches, which combined the image-based nature of DOS with the precision of an RTC, provided sufficient accuracy and simplicity to warrant consideration for use on an operational basis. Though these results were probably site-specific (characterized by relatively low aerosol levels and low humidity), they illustrate the feasibility of simple atmospheric correction methods and the usefulness of a diverse data set for validation of such techniques.

An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data

Digital analysis of remotely sensed data has become an important component of many earth-science studies. These data are often processed through a set of preprocessing or “clean-up” routines that includes a correction for atmospheric scattering, often called haze. Various methods to correct or remove the additive haze component have been developed, including the widely used dark-object subtraction technique. A problem with most of these methods is that the haze values for each spectral band are selected independently. This can create problems because atmospheric scattering is highly wavelength-dependent in the visible part of the electromagnetic spectrum and the scattering values are correlated with each other. Therefore, multispectral data such as from the Landsat Thematic Mapper and Multispectral Scanner must be corrected with haze values that are spectral band dependent. An improved dark-object subtraction technique is demonstrated that allows the user to select a relative atmospheric scattering model to predict the haze values for all the spectral bands from a selected starting band haze value. The improved method normalizes the predicted haze values for the different gain and offset parameters used by the imaging system. Examples of haze value differences between the old and improved methods for Thematic Mapper Bands 1, 2, 3, 4, 5, and 7 are 40.0, 13.0, 12.0, 8.0, 5.0, and 2.0 vs. 40.0, 13.2, 8.9, 4.9, 16.7, and 3.3, respectively, using a relative scattering model of a clear atmosphere. In one Landsat multispectral scanner image the haze value differences for Bands 4, 5, 6, and 7 were 30.0, 50.0, 50.0, and 40.0 for the old method vs. 30.0, 34.4, 43.6, and 6.4 for the new method using a relative scattering model of a hazy atmosphere.